[sdiy] Speaking of the Elektor Vocoder (and the Korg Vocoder)

[anthony]

> On Wed, February 20, 2008 12:05 pm, Eric Brombaugh said:
>
>> My point was that if sample rates are high enough (and hardware is
[snip]

> It's still present, even if it's not noticeable on the human scale. That
> was all I took Anthony's original comment to mean - that he was noting in
> passing that he was comparing an "instantaneous approximate Fourier
> transform" in the analogue world to a fast Fourier transform in the
> digital world. IE different math with similar results.
>

That is exactly what I was saying. If you had 100 or even 1000 fixed 
badnpass filters with high enough Q's and had bar graph meters or 
what-have-you on each one, the result would be a spectral analysis much like 
what the Fast Fourier Transform algorithm would yield. But the way you would 
mathematically express the different mehods is, well, different (that's as 
far as I'll go because I don't actually consider myself t be a "math guy"). 
The analog advantage is that each band analysis is happening at exactly the 
same time: great for rapidly changing spectral information.
The disadvantage would be the number of high Q filters you would need and a 
basically fixed resoltuion. Although I suppose you could have a way to use 
all of the banks to analyze frequencies very close to each other - which 
would of course require filters of even higher Q (something like -90dB).

I think if you did something like heterodyning a swept oscillator over a 
swept filter, this might have math behind it that is more similar to the FFT 
algorithm. But I don't think you could use this method to do anything like a 
vocoder. And accuracy and resolution would be abysmal unles you build some 
pretty complex and precise gear. This kind of thing is a piece of cake for 
the RF world - especially when you're looking at something like sidebands on 
an AM transmitter. (Oh yeah - I'm remebering a project to turn an 
oscilloscope to a spectrum analyzer.)

And true: in the digital world, having a lot of processor power to throw at 
the FFT, you can get a reponse as close to instantaneous as you could 
probably keep track of with your eyes and then some (unless it's some stupid 
visualizer in Windows Media Player...)